메뉴
How to run Sprint 0 for a new Rust project in this squad
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
| name | sprint-zero-planning |
| description | How to run Sprint 0 for a new Rust project in this squad |
| domain | project-setup |
| confidence | high |
| source | earned |
Sprint 0 runs before any implementation begins. Its job is to establish structure, gates, and ownership so all parallel workstreams (impl, test, docs, review) stay coordinated from the first commit.
Read the spec first — understand phases, ship gates, and reviewer requirements before decomposing.
Create OpenSpec proposals for each major phase before touching code. Proposals live in openspec/changes/<name>/proposal.md. They must include: scope, non-goals, ship gates, and reviewer assignments.
Scaffold structure matches the spec's Repository Structure section exactly. Stubs use todo!() — no logic, just the right signatures and module tree.
Schema is spec, not implementation — src/schema.sql should be created in Sprint 0 from the DDL in the spec. It's the contract everything else builds to.
CI gates from day one — ci.yml should run on every PR: cargo fmt, cargo clippy, cargo check, cargo test, static binary verification. Don't wait for Phase 3 to add these.
Phase gates must be explicit — each OpenSpec proposal's depends_on and Ship Gate section enforce sequencing. No Phase 2 proposal should be actionable until Phase 1 gates pass.
CLAUDE.md and AGENTS.md belong in Sprint 0 — any agent spawned in the repo needs to orient quickly. These files should be created before implementation starts.
See openspec/changes/sprint-0-repo-scaffold/proposal.md for a complete Sprint 0 proposal.
src/schema.sql — without it, db.rs has no contract to implement against.create tool to create directories — it requires parent dirs to exist. Use a general-purpose agent with Python to create directory trees first.pwsh.exe (PowerShell 7) is NOT available. Use a general-purpose agent with Python to create directory trees.Resolve knowledge gaps: fetch unresolved gaps, assess sensitivity, generate research queries, ingest findings, and close gaps with a resolution slug.
Interrupt-driven alerts: detect and surface new contradictions, stale pages, resolved gaps, and embedding drift. Priority-classified with deduplication.
Ingest meeting notes, articles, documents, and conversations into Quaid. Handles idempotent ingestion for exact-byte duplicates and vault-backed sync.
Keep the brain healthy: detect and resolve contradictions via the correction / supersede workflow, validate referential integrity, triage knowledge gaps, find orphaned pages, and compact the database.
Answer questions from the brain using FTS5 + semantic search + structured queries. Synthesize across multiple pages. Cite sources.
Agent-guided binary upgrades: version check, channel detection, download via install.sh (or manual asset fetch + SHA-256 verify), and post-upgrade validation.